Design and Demonstration of an Interference Suppressing Microwave Radiometer Grant A. Hampson, Steven W. Ellingson*, and Joel T. Johnson 2004 IEEE Aerospace Conference

Department of Electrical and Computer Engineering ElectroScience Laboratory The Ohio State University

ElectroScience Lab

*Department of Electrical and Computer Engineering Virginia Tech 10 th March 2004

RFI Issues for Microwave Radiometers

 A microwave radiometer is a sensitive receiver measuring naturally emitted thermal noise power within a specified bandwidth  Human transmission in many bands is prohibited by international agreement; these are the “quiet bands” ideal for radiometry  L-band channel quiet band is 1400-1427 MHz: larger bandwidth would improve sensitivity if RFI can be addressed. No quiet band available for C-band radiometers  Even within quiet bands, RFI has still been observed - possibly due to filter limitations or intermodulation products  Many interferers are localized either in time or frequency: should be relatively easy to detect and remove with an appropriate system

ElectroScience Lab

Outline

 Design of an RFI suppressing radiometer  Description of our implementation  Experiment description  Initial results

ElectroScience Lab

System Overview

     Typical radiometer is a very “slow” instrument: power received is integrated up to msec scales by analog system before being digitized Typical radiometer has a single, large bandwidth channel: susceptible to narrow band interference Our design uses a digital receiver to allow much more rapid sampling of incoming data; this rapid sampling improves the ability to mitigate temporally localized RFI Our design also performs a 1024 point FFT operation; improves ability to mitigate spectrally localized RFI Processor must operate in real time so that final data rate can be reduced to a manageable level; implement processor in hardware (FPGA’s) Pulsed interferer (~ m sec) Radiometer integration period (~msec) Time

ElectroScience Lab

System Block Diagram

Antenna Low-noise front end Analog Downconverter ADC Digital Downconverter Asynchronous Pulse Blanker 1024 point FFT Frequency domain blanker (not yet implemented) Integration Data Recording/ Control

ElectroScience Lab

Example: Time Blanking of an ATC Radar

 Time domain results: Direct path Multi-path?

 APB “Blanking” decision Effect of varying APB threshold in frequency domain: “Max held” spectra Averaged spectra

ElectroScience Lab

Digital Back-End

 System design includes digital IF downconverter (DIF), asynchronous pulse blanker (APB), FFT stage, and SDP operations  Analog Devices 9410 ADC ADC DIF APB 200 MSPS 100 MSPS I/Q Dual channel implementation completed FFT SDP  Most blocks on separate boards to simplify testing and reconfiguration  Microcontroller interface via ethernet for setting on-chip parameters  Second prototype uses Altera "Stratix" FPGA’s: apprx 10000 LE, $260  Designs for all components complete; DIF, APB, FFT, SDP, and capture card initial implementations functioning

ElectroScience Lab

Current Digital Back-End Implementation

  Modular form used for processor boards: note microcontrollers EEPROM's on each card for autoprogramming of FPGA's on power-up ADC DIF/ APB FFT SDP Capture ADC

ElectroScience Lab

Experiments at OSU

    Experiments designed to demonstrate radiometric accuracy in the presence of interference Observations of a large water tank; external cal sources are ambient absorbers and a sky reflector Highly accurate ground-based radiometry is tough due to contributions from objects not under view, including reflections Keep cal targets exactly the same size as pool to reduce these effects; observations of pool as ambient temp varies also  Initial tests in existing RFI; artificial RFI to be added as tests progress

ElectroScience Lab

Antenna/Front End Unit

 Front end Tsys approx. 200K neglecting antenna Temperature control used to maintain internal standards; rest of system not temperature controlled Terminator total power within 0.01 dB over ~1 hour after noise diode stabilization

ElectroScience Lab

Dual Channel Downconverter

 One channel is ~1325-1375 MHz, other is ~1375-1425 MHz   Downconverter, digital receiver, computer, and thermal control systems in rack inside lab Working with system gain ~70 dB at present: moves –94 dBm in 100 MHz to –24 dBm; this triggers 4-5 ADC bits

ElectroScience Lab

  

Pool and Cal Targets

Water pool is approximately 16’ x 32’ by 6” deep; use saline water to decrease skin depth and eliminate sub-pool contributions Absorbers are 12” (a few 18”) to cover the area; mounted on “racks” that can be placed on pool by a team of 2-4 people Reflectors are foil covered foam  Temperature of pool, absorbers, and ground around pool all recorded; salinity of pool checked for each experiment

ElectroScience Lab

Pool and Cal Targets

Absorbers: Assume Tb=Tphys Reflectors: Assume Tb=Tref~Tsky?

Possible background contributions make precise calibration tough; data analysis to sort out these effects currently in progress Water: Tb~Twat+QTref

ElectroScience Lab

Current Experiment Results

 Hardware integration period 21 msec, switched every 1.33 secs through front end states, every 5 seconds through receiver mode Horizontally polarized data, absorber observation Radar at 1331 MHz apparent, along with other more CW RFI Blanker reduces radar effects, as well as saturation problems in FFT computation

ElectroScience Lab

Initial Calibrated Brightnesses

 Water observation using absorber/reflector cal, horizontal pol Blanker dramatically reduces influence of radar at 1331 MHz CW RFI in band not affected by blanker Some spectral features remain in blanker on data; caused by calibration issues Absolute value near expected water brightness Working to improve calibration/accuracy of this experiment

ElectroScience Lab

Total Power Measurement Vs. Time

 Water observation using absorber/reflector cal, horizontal pol Radar causes periodic huge change in total brightness; also a calibration bias through effect on absorber/reflector Blanker greatly reduces these effects Frequency domain blanker implemented in software; simple threshhold of brightness data Narrowband RFI not a major impact on total power in this experiment

ElectroScience Lab

Summary

 Digital receivers have the potential to greatly improve radiometer performance in the presence of RFI: essentially a using a more capable receiver  Many options available beyond those chosen in our implementation; our basic algorithms still show reasonable initial performance  Still working to refine demonstration; improve accuracy to be able to demonstrate suppression of smaller RFI effects  We have also operated an L-band RFI monitoring system in aircraft experiments to gather data in a wider range of environments  Our system will be operated in C-band aircraft experiments in 2004 to explore suppression performance and to gather C-band RFI environment data

ElectroScience Lab